List your company

Step-wise Verification on ProcessBench Math

69.5TNRCons@4
12.327.154256.85May 20, 2026

Evaluation results

MethodLinks
Cons@4Backbone=Qwen3-8B2026.0569.586.377
VerifySteer-UniBackbone=Qwen3-8B2026.056985.576.4
VerifySteer-BiBackbone=Qwen3-8B2026.0568.786.976.7
Qwen3-14BBackbone=Qwen3-14B2026.0568.586.976.6
Prompt Opt.Backbone=Qwen3-8B2026.0566.88574.8
Basic PromptBackbone=Qwen3-8B2026.0558.48368.6
Llama-3.3-70BBackbone=Llama-3.3-70B2026.054694.361.8
Qwen2.5-72BBackbone=Qwen2.5-72B2026.0545.891.461
VerifySteer-UniBackbone=Qwen3-1.7B2026.0521.761.632.1
VerifySteer-BiBackbone=Qwen3-1.7B2026.0521.569.232.9
Cons@4Backbone=Qwen3-1.7B2026.0520.369.431.4
Prompt Opt.Backbone=Qwen3-1.7B2026.051858.427.5
Basic PromptBackbone=Qwen3-1.7B2026.0514.584.224.7